Recommendation System using Inference-based Graph Learning – Modeling and Analysis

Recently, the development of recommendation systems has been an emerging area where the target is the learner's preferences, and styles should be learned from past interactions to predict the ratings and recommendations of the new items. The significant critiques and challenges are some of the datasets are sparse and the bias in the ratings and popularity of the items that conventional recommenders can't efficiently process. This research constructs the graph model to merge the items and user interaction to enhance user representations. The new inference operator is designed to remove the bias factors and to draw inferences. The users and items are represented with high quality by constructing the semantic space that connects the users and the item's characteristics. The proposed model is simulated on real-world datasets, proving that the model produces significant outcomes better than existing methods.